Prediction of nanostructure of SnO2 properties using artificial neural networks / Khadijah Mohd Suhami ... [et al.]
Tin(IV) oxide, SnO2 nanostructures such as nanorods, nanoflowers, nanosheets, nanocubes have been receiving significant interest in various fields due to their inherent properties. The types of shape and size of nanorods vary based on the applications. A fine tuning of the parameters ( e.g concentra...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Conference or Workshop Item |
| Language: | English |
| Published: |
2021
|
| Subjects: | |
| Online Access: | https://ir.uitm.edu.my/id/eprint/56910/1/56910.pdf https://ir.uitm.edu.my/id/eprint/56910/ https://ispike2021.uitm.edu.my/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Tin(IV) oxide, SnO2 nanostructures such as nanorods, nanoflowers, nanosheets, nanocubes have been receiving significant interest in various fields due to their inherent properties. The types of shape and size of nanorods vary based on the applications. A fine tuning of the parameters ( e.g concentration, pH, temperature, template, type of solvent etc.) during the synthesis process can alter the morphology of the SnO2. However, producing nanostructures with the desired size and shape is extremely complex and still remains a challenge. Hence, in this study a mathematics modelling called Artificial Neural Network (ANN) for the prediction of the SnO2 morphology was developed. This study was carried out using the real time data collected via experimental work and training the data using a neural network toolbox in MATLAB Version (R2016a) software. An ANN modelling was constructed with the input parameters of reaction time and concentration of precursors and three different output parameters namely, crystalline size, band gap energy and size of particles. This modelling was developed based on trial and error at different network architecture, activation function and training algorithm. The data set was trained using hyperbolic tangent sigmoid (tansig) activation function and Levenberg-Marquardt training algorithm. The performance of modelling was evaluated based on the mean square error (MSE)
and coefficient of determination (R2). The finding shows, there is no overfitting while constructing the neural network and it is able to track the data. The result shows that the MSE performance plot and R2 are in the range of 0.1-1.0. Therefore, it is suggested that the ANN modellings constructed in this study are able to produce a decent prediction. These values indicate that prediction of nanostructure SnO2 properties using artificial neural network (ANN) is a great success. |
|---|